This code is for the paper Bitcoin Anti-Money Laundering Based on BGNN Graph Neural Network
The BGNN Graph Neural Network algorithm is used in the project which is from the ICLR 2021 paper: Boost then Convolve: Gradient Boosting Meets Graph Neural Networks
The elliptic dataset is from the kaggle.
This code contains implementation of the following models for graphs based on the bgnn model:
- BGNN (end-to-end {CatBoost + {ARMA, hGANet, DAGNN}})
First, you should download the repository and install some necessary packages according to the requirements, you can use the following command:
pip install -r requirements.txt
1) first you should process the original dataset, generate the graph dataset, you can execute the following code:
python scripts/preprocessing_elliptic.py
2) train the test the model.
nohup python scripts/run.py datasets/elliptic bgnn --max_seeds 5 \
--repeat_exp 5 --task classification \
--metric f1 --save_folder ./results/elliptic_bgnn \
--version_num 1.0 2>&1 > log_test_202109.txt &
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent